Response Time Analyzer

Server assisted Performance

Sends a small, rate-limited series of requests (max 20) to a public URL and reports latency statistics, TTFB and consistency, with a simple chart.

Privacy: This tool sends the request through the Mutqan backend relay to reach the target. Only the target URL, headers and body you enter are transmitted, and nothing is stored.

How to use Response Time Analyzer

  1. Enter a public http(s) URL (private and loopback addresses are refused by the relay).
  2. Choose the number of samples (1–20) and the method (GET, HEAD or OPTIONS).
  3. Click "Measure" and read the statistics: the median describes typical latency, p95/p99 the tail.
  4. Check the consistency score and the chart for outliers before comparing runs.
  5. Download the samples as CSV or run again to compare before/after a change.

Response Time Analyzer features

  • Sends 1–20 sequential requests through the Mutqan relay and records every sample
  • Min, max, average, median, p90, p95, p99 and standard deviation
  • Consistency score that highlights jitter and cold-start outliers
  • Bar chart of individual samples plus a latency distribution table
  • Status breakdown so failed or rate-limited requests are never hidden in the average
  • Downloadable CSV of the raw samples for reports and dashboards

Response Time Analyzer example

Ten samples against a health endpoint

Input:

URL: https://api.example.com/health
Samples: 10

Output:

Samples: 10
Min: 112 ms · Average: 138 ms · Median: 131 ms
p90: 171 ms · p95: 182 ms · p99: 190 ms · Max: 190 ms
Std deviation: 24 ms (17% of average)
Consistency: 78/100 — stable
Statuses: 200 ×10

Frequently asked questions about Response Time Analyzer

How many requests does the analyzer send?

Between 1 and 20 sequential requests through the Mutqan relay, with a timeout per request and per-IP rate limits. It is designed for quick health checks, not sustained load.

What do p90, p95 and p99 mean?

They are percentiles: p95 = 95% of the samples were at least this fast. Percentiles expose tail latency that averages hide, which is why SLOs are usually written as p95 or p99 targets.

Why are my numbers higher than in my own monitoring?

Timings include the network path from the relay to your origin. Compare relative values (before/after a change, endpoint A vs B) rather than absolute milliseconds.

What is the consistency score?

A 0–100 score derived from the coefficient of variation (std deviation ÷ average) and the p95/median ratio. Low scores mean jitter — look for cold starts, caching misses or noisy upstream calls.

Can I test private or localhost endpoints?

No. The relay refuses private, loopback and link-local addresses. Use it only for public URLs you are allowed to test.

Technical notes

Percentiles use linear interpolation between sorted samples, matching the definition used by k6 and most APM products. With a small number of samples p99 is close to the maximum; treat it as an indication rather than a guarantee.

The consistency score blends the coefficient of variation (standard deviation divided by average) with the p95/median ratio. A perfectly stable endpoint scores 100; a p95 three times the median scores 0 on the tail component.